Account-Based Marketing Success Stories: How to Read Them

Demandbase's homepage carries a four-tile band of account-based marketing case study results, each tile a logo, a number, and a link to that customer's case study. Two of the four cite a figure that appears nowhere on the page they link to. The SAP Concur tile shows a 3× increase in conversions, which is the multiple on Adobe's case study, while SAP Concur's own page reports 4X; the Adobe tile shows a 52% increase in revenue that the Adobe case study never mentions. The Thermo Fisher and Ingram Micro tiles, +50% average deal size and +83% pipeline velocity, match their pages (Demandbase homepage).

That is not a reason to dismiss an account-based marketing case study. It is a reason to read one with a method. Eleven published account-based marketing success stories were read for this piece from the vendors' own pages, quoted from the served bytes rather than from a summary, and every claim below is that vendor's claim about its own customer, attributed to the page it sits on.

The pattern across all eleven is consistent enough to be useful. The numbers are real in the sense that someone published them. What is missing is the same set of things every time, and once you know which things, you can read any ABM success story in about two minutes and know how much weight it will carry.

Account-based marketing case studies: the eleven, as published

Every account-based marketing case study in this set is a vendor's own account of a customer result: five from Demandbase, two from DemandScience, three from AdRoll ABM and one from LinkedIn Marketing Solutions. The headline figure for each is quoted below as that vendor's page states it, and every source page is linked at the foot, so any row can be checked against its page in a minute.

Read them yourself if you are evaluating any of these platforms, because the summaries above are the vendors' framing and the interesting material is underneath it.

Reading a vendor's own page for what it states rather than what it implies also works on pricing, where two published Unbounce boundaries decide which alternatives are worth a shortlist.

Six questions that separate a usable story from a decorated one

The six questions are: is there a before number for the headline metric, over what period was it measured, what did the programme cost, what is the comparison group and who chose it, who is saying the number, and is the metric defined and measured by whom.

Three of the six are answered sometimes across the eleven. Three are answered rarely or never.

Cost is answered zero times out of eleven. Not one of the eleven pages states a licence fee, a contract value, a seat count in money, a media budget, or a cost per lead in currency. Quit Genius's page is the sharpest example: it says lead volume and cost per lead were "guaranteed up front" and then declines to say what either number was, which is the single most checkable fact the story could have carried (DemandScience). A return figure with the investment side removed is a gross number, and every one of these is a gross number.

A control group appears zero times out of eleven. Every comparison is the customer against their own prior state, or exposed accounts against unexposed ones where exposure was chosen by an engagement filter. PitchBook's page states its filter in writing: to count as influenced, "an account had to have at least 15 ad impressions and either 1 ad click or 1 conversion" (AdRoll). Accounts that click your ads fifteen times are accounts already in market. The claim that influenced deals close bigger and faster is then substantially a restatement of the claim that engaged accounts buy more, and the page presents the filter as rigour.

A named customer is quoted saying a number in two cases out of eleven. Snowflake's Hillary Carpio is quoted with "We're achieving a 50 percent new opportunity rate with existing customers we target with ABM", a figure that appears in no tile on the page and is defined nowhere. Navisite's Matt Norris is quoted with rep adoption rates. Total Expert's page is a third shape: it is written in the first person under a named manager's byline, so its 100% figure arrives in the customer's voice without a quotation. In the other eight, any customer quote is qualitative and every figure is asserted in the vendor's own narrative voice. That distinction matters more than it looks: a marketing team can publish a number its customer never said out loud.

The best-disclosed story and the least, side by side

SAP Concur's page earns its numbers. It states that time in the engaged stage had been 137 days and that those visitors now convert to the next stage in 35 days, which is a real before and after (Demandbase). Two caveats survive even so. The arithmetic of 137 to 35 days is 3.91 times, rounded up and printed as 4X. And the segment being measured is defined by high-intent behaviour, repeat visits and video views and paid-search arrivals, so faster progression by accounts selected for already showing buying behaviour is close to circular. The page says as much without noticing, describing the exercise as having "proved their hypothesis".

Thermo Fisher's page is the specimen at the other end. Three headline figures sit in tiles, the narrative that is supposed to explain them mentions none of them, and a search of the whole document for a year, month or duration returns only the footer copyright (Demandbase). The page quotes no customer and names no employee. An 18% increase in revenue at an 80,000-employee company would be enormous in absolute terms and is almost certainly scoped to something far narrower, but the page never says which population the percentage covers.

Four specific failures worth learning to spot

A percentage attached to the opposite metric. Ingram Micro's headline is an 83% increase in pipeline velocity, and the supporting line is "reducing their sales cycle of 12 to 2 months" (Demandbase). Twelve months to two is an 83% reduction in cycle length. Expressed as velocity, the reciprocal, that is a six-fold increase. The 83% belongs to the metric that went down and has been printed against the metric that went up. The correct velocity figure would be far larger, so this is not a vendor inflating a number. It is a unit error nobody caught, which tells you how carefully the figures are produced.

A coverage statistic dressed as a performance one. Total Expert's page leads with "This year, 100% of our closed-won deals carried AdRoll ABM attribution" (AdRoll). If you advertise to and contact your entire target account list, every deal will carry a contact. The page then explains how the denominator got there: deals used to close without attribution, and the fix was to redefine the ICP so sales stopped pursuing accounts outside it. The metric moved because the population was redefined. There is also no deal count anywhere on the page, so 100% of closed-won deals could be three deals.

Two numbers that cannot be checked against each other. Cardinal Health WaveMark's tile reads 417% growth in pipeline, while the body gives the same 417% as the increase in marketing-influenced opportunities, and separately says 10% of opportunities were marketing-influenced in FY 2023 against 68% in FY 2024 (DemandScience). On the page's own figures, a 417% increase is a 5.17-fold count, while 10% to 68% is a 6.8-fold share. Both hold at once only if total opportunities fell by about a quarter, and the page never gives the total. Worth noting the same page carries the best metric definition in the set: the customer defines marketing-influenced in her own words as an opportunity that "was either part of a recent engagement spike or had clicked on our advertising, so not just views or impressions, but actual clicks".

A page whose title names a different customer. The Quit Genius case study sits at a Quit Genius address under a Quit Genius heading, but its title tag reads "PureSyndication and PureABM Accelerate CircleCI Sales Cycle" (DemandScience). A reader of the body never sees it; a browser tab does. It is a small error, and the kind that says nobody re-reads these pages after they ship.

Publication dates tell you nothing

PitchBook's and Snowflake's case studies carry an identical datePublished of 2025-12-29 in their structured data, on stories describing very different periods. Snowflake's narrative is set around a global pandemic and the company's IPO. That shared timestamp is a content-management migration stamp, not a story date. When a case study offers no measurement window in its copy, the publication date will not rescue it.

What a story you could actually act on would contain

Ask for five things: a reference call with the named customer, asking them the before number directly; the denominator, meaning how many accounts were in the programme and how many deals sit behind the percentage; the window, with start date, end date and a baseline period of the same length; the cost line, meaning licence, seats and media for the period the result covers; and the comparison, meaning who was left out of the programme and what happened to them.

None of those are unreasonable questions and a vendor with a good story can answer all five on a call. The reason to ask them in that order is that the first two usually settle it. A customer who cannot state their own before number was not measuring, and a percentage with no denominator behind it survives no scrutiny at all.

If you are earlier than that, and you are still working out whether an account-based programme is the right shape for your pipeline problem at all, the sensible order is to settle the target list and the qualifying situation before you evaluate anyone's platform. Our own view on when account-based marketing beats broad outbound sets out the situations where it earns its cost, and the account-based marketing strategy piece covers building a target list you can work. For the vendor question specifically, ABM platforms compared and the Demandbase review both start from what each product replaces rather than from its case studies.

How we treat this on our own work

We hold ourselves to the same six questions, which is why we publish a defined qualified meeting standard before a campaign launches rather than a percentage afterwards. Meetings are qualified against criteria agreed in writing before anything sends, and budget, timing and authority are never conditions of billing. If you want to see the shape of that in practice, the case studies carry the definitions alongside the numbers, and a free campaign is scoped against the same written criteria before it runs.

The reason for the discipline is not modesty. A result you cannot state the baseline, period and cost for is a result you cannot repeat, and the client cannot tell whether it worked.

The short version

Eleven published ABM success stories, read from the vendors' own pages: none discloses what the programme cost, none carries a control group, two quote a named customer saying a number, and the strongest one still omits its sample size and measurement window. Two of four tiles on Demandbase's homepage cite figures absent from the case studies they link to. Read any of these stories by asking for the before number, the period, the cost, the comparison group, the speaker and the metric definition. Most published ABM results answer two of those six, and the two they answer are rarely the ones that decide anything.

Vendor claims, figures and page content are taken from each vendor's own pages, from stored snapshots of the served bytes. Every figure above is the vendor's claim about its customer, not an independent measurement. Verify current terms and current published claims with the vendor before relying on them.

Sources: Demandbase homepage, Adobe case study, SAP Concur case study, Thermo Fisher case study, Ingram Micro case study, Navisite case study, Quit Genius case study, Cardinal Health WaveMark case study, PitchBook case study, Snowflake case study, Total Expert case study, Refinitiv customer story.